Facial Recognition for Fish: The New Frontier of Precision Aquaculture

China, 12 November 2025 | FishFaceID uses artificial intelligence to revolutionizing the monitoring and management of farmed aquatic animals

Holoturia (pepino de mar). Universidad de Stirling

Unlike traditional livestock farmers, fish farmers cannot accurately account for the number of animals they have underwater or identify them individually. However, technological advances are rapidly closing this gap.

A research team from the Ocean University of China has developed a non-invasive identification system based on deep learning, capable of performing individual tracking of aquaculture organisms without the need for physical tags or marks that can cause stress, slow growth, or increase vulnerability to disease.

The tool, named FishFaceID, enables the individualized management of the fish, opening the door to personalised feeding regimes, more precise genetic selection, and improved health prevention – all key pillars of modern Precision Aquaculture.

Crucially, the system employs an innovative deep learning model called Vim-FFD, based on the Vision Mamba architecture, which is specifically designed for fine-grained underwater recognition. According to its authors, this approach offers high accuracy and robustness even in the face of adverse conditions typical of commercial farms, such as water turbidity, individual overlap (occlusion), or changes lighting and camera setting between sessions.

The researchers report a striking 99.81% accuracy (Acc@1) on the sea cucumber subset, demonstrating the method’s clear visibility.

In a departure from previous studies, which were confined to a single species or controlled laboratory environments, this work proposes a scalable solution for the industry. It has been validated across a multi-species benchmark, including both marine and freshwater organisms such as the sea cucumber, leopard coral grouper, blue-spotted grouper, and grass carp.

Furthemore, the Qindao and Sanya-based team has released the generated dataset publicly, aiming to foster the development of open and collaborative technologies for next-generation aquaculture.

Related